HJB Optimal Feedback Control with Deep Differential Value Functions and Action Constraints

no code implementations • 13 Sep 2019 • Michael Lutter, Boris Belousov, Kim Listmann, Debora Clever, Jan Peters

The corresponding optimal value function is learned end-to-end by embedding a deep differential network in the Hamilton-Jacobi-Bellmann differential equation and minimizing the error of this equality while simultaneously decreasing the discounting from short- to far-sighted to enable the learning.

reinforcement-learning Reinforcement Learning (RL)

Deep Lagrangian Networks for end-to-end learning of energy-based control for under-actuated systems

1 code implementation • 10 Jul 2019 • Michael Lutter, Kim Listmann, Jan Peters

Applying Deep Learning to control has a lot of potential for enabling the intelligent design of robot control laws.